ABSTRACT Pancreatic cancer is an almost uniformly lethal disease. This year in the United States, an estimated 37,680 patients will be diagnosed with pancreatic cancer, and 34,290 patients will die of their disease. In contrast to the wealth of data demonstrating genetic or epigenetic alterations associated with pancreatic carcinogenesis, the genetic events specifically and consistently associated with pancreatic cancer metastasis have not been well defined. However, it is this final stage of the cancer process- the uncontrolled growth and/or dissemination of cancer cells to other organs-that is ultimately responsible for the majority of cancer-related deaths. This fact is of particular relevance to human pancreatic cancer. The dismal prognosis associated with this disease is largely due to the fact that most patients are diagnosed at an advanced stage of disease that is not amenable to surgical resection. Studies that focus on pancreatic cancer metastasis have the potential to greatly expand our understanding of the most lethal stage of this disease and identify novel areas for therapeutic intervention. We have recently characterized the patterns of pancreatic cancer failure in rapid autopsy participants and found that genetic inactivation of the DPC4 gene is highly correlated with widespread metastatic failure at autopsy. Thus, we propose three Specific Aims towards understanding the contribution of the TGF- pathway to pancreatic cancer progression. First, we will perform high throughput sequencing and copy number evaluations of matched primary and metastatic tissues from 48 patients who have undergone rapid autopsy, and use this unprecedented dataset to identify the pro-metastatic genotype of pancreatic cancers. This genotype classifier will be independently validated in an independent set of early stage pancreatic cancer tissues from patients with known outcome. Second, we will determine the synergistic effects of TP53 and TGF- pathway alterations in pancreatic carcinogenesis and progression using two well characterized mouse models of pancreatic cancer. Third, we will use well characterized cell lines derived from rapid autopsy patients to determine the effects of DPC4 restoration, or DPC4 knockout, on both canonical and noncanonical TGF- pathway signaling and their relationship to invasion and spontaneous metastasis in vivo. These findings will be important because they will lead to improvements in therapeutic management of patients with pancreatic cancer based on their expected patterns of failure, including the development of novel rational therapies for this disease.